Dan Sperber in Cognition and Culture:
Your friend Olga is coming for a drink. You put two plates on the table, one with olives and the other with almonds. When both plates have been emptied, you ask Olga, “Do you want anything else?” “Yes, please!”, she answers, pointing to the plate where the almonds had been. What is she requesting? The empty plate? Of course not. She is requesting more almonds. To do so, she uses a gestural metonymy: pointing to a container to convey something about its (past) content.
Container-for-content metonymies are quite common in language use. Typical examples are: “I just had one glass” or “the school bus was singing.” Some of these gestural or verbal metonymies have become conventional but we can produce or understand novel ones without effort. What communicative abilities does it take to make use of metonymies? Could a 12-month-old child, who does not yet speak, spontaneously produce an appropriate gestural metonymy? For that matter, could an ape?
In his doctoral work at the Max Planck Institute for Evolutionary Anthropology, Manuel Bohn asked an even more basic question: can infants and apes refer to absent entities? (See also earlier work by Liszkowski et al 2009; Lyn et al 2014). The capacity to do so is generally linked to the possession of language, so showing that they can would be an interesting challenge.
In one study (Bohn et al 2015), Bohn presented apes (chimpanzees, bonobos, gorillas, and orangutans) with two plates each containing three pieces of food: grapes (a higher quality food for apes) on one plate, and pieces of carrot (a lower quality food) on the other plate. The apes could point to one or the other plate and would be given a piece of food from it. As soon as a plate was emptied, the experimenter would take it out of the room and bring it back refilled.
In the critical test trials, however, the experimenter let the plates go empty without refilling them. Would the apes point to a now empty plate (as your friend Olga did)?
More here.